On the face of it, the answer to this question should presumably be resoundingly affirmative. With far more efficient utilisation of available compute resources, economies of scale in staffing and infrastructure, and a tendency to site the bigger data centres in areas blessed with natural cooling and abundant green-ish power, the answer is surely obvious. How could a Cloud be any less Green than all those on-premise data centres, full of woefully under-utilised computers that endlessly pump waste heat toward a roof filled to overflowing with heat exchangers and power-guzzling air conditioning units?
Surely, as world leaders fall over themselves to exhort someone else to make the painful decisions ahead of Copenhagen, Cloud Computing is the technology community’s contribution to making sure the world is still here for our children to inherit and enjoy?
Chris Thorman has an interesting post that is unusual in actually trying to apply some real numbers to the discussion. In it, he looks quite specifically at the direct energy costs of running a particular medical application on-premise or hosted by a third party;
“On-premise software is what most people think of when they think of a software system. You pop in a CD or download a big file from the Internet and the install begins. Files are copied to your computer and/or a server machine, where they are stored and run. Because the client and server software components are both doing a lot of computations, a fair amount of power is required.
With SaaS, there is no local installation of software because the vendor manages all the code and the data in their data center. Users access the system through a web browser and its primary role is to present the user interface – not a very computationally intensive function.”
The conclusion, at least based upon the factors Chris chose to measure, is unsurprising;
“That’s an 93% reduction in overall energy consumption for a four physician practice using SaaS EMR software over on-premise software!”
Simon Wardley, one of the Cloud’s more effusive cheerleaders, offers a less straightforward analysis in a recent video conversation with GreenMonk‘s Tom Raftery. It’s well worth watching, although you may join me towards the end in wishing that the fancy fades would stop…
Unsurprisingly, Simon agrees with my opening assertions that compute resources in the Cloud should be more efficiently utilised and therefore cheaper to run on a per-unit basis. However, Simon goes on to suggest that the near-ubiquitous availability of computation will lead to a profligate rise in usage. We might, as Chris Thorman suggested, complete a specific task more cheaply and efficiently than before, but we’ll run more tasks without giving sufficient thought to the consequences. Just as lights and gadgets get left on all over the house because the power is just there (and too cheap to worry about in most cases), so too will ‘available’ computers be squandered. It will be easy to call upon a computer when required, and application developers will be quick to spot the opportunities to put those spare cycles to work in completing myriad small, individually inconsequential yet collectively shockingly wasteful tasks. We see this with desktop computers today, casually devoting once-scarce resources to the generation of ‘eye candy.’ On an individual basis the ‘waste’ is small and easily ignored. Scaled to the level of the Cloud it will surely become far more noticeable, far more measurable, and far less easy to justify.
Is this the Cloud’s ‘fault?’ No, of course not. Instead it appears baked into the attitudes with which we view commodities. So whilst the technologies behind the Cloud might make Green efficiencies possible, human nature may very well pull in exactly the opposite direction. Which big Cloud provider will be brave enough to implement a ‘Green tax,’ charging customers a premium for ‘wasteful’ computing? And who will decide what’s wasteful, anyway?
And Simon’s final answer to the question, ‘is Cloud Computing Green or not?’
“Yes and No.”
In related news, I’ve been reading the latest installation of Rackspace‘s Green Survey. When directly asked,
“Do you view cloud computing as a greener alternative to traditional computing infrastructures?”
- 21% agreed, “Yes, cloud computing is a much greener alternative”
- 35% were “not convinced of the green benefits of cloud computing”
- 25% reckoned that there was “too much hype around the green benefits of cloud computing”
- 19% suggested that “the true green benefits of cloud cmputing have not yet been realized.”
Asked in a subsequent question about how cloud computing fits within the organisation’s broader set of environmental initiatives, 7% claimed that “cloud computing is critical to our company becoming greener.”
Respondents were all Rackspace customers, and I’m actually surprised that the pro-Green numbers weren’t higher. Did respondents not understand the efficiency argument, did they not believe it, or had they moved past it to embrace Simon’s arguments about our penchant to waste anything abundant?
There are some fascinating figures in the survey, although like many similar exercises I’m left with as many (different) questions as when I started.
So… Are Clouds green?
I think I’ll follow Simon in hedging with ‘Yes and No.’
What do you think?
Related articles by Zemanta
- IBM data center gets energy overhaul (news.zdnet.com)
- Using the Sea To Cool Your Data Center (hardware.slashdot.org)
- Novel way to cool datacenters passes first test (infoworld.com)
- IBM Pushing Water-Cooled Servers, Meeting Resistance (hardware.slashdot.org)
- The Hidden Cost of the Cloud: Bandwidth Charges (gigaom.com)